audio pre-amp design (long-ish)

So, I've been working on an audio pre-amp design for a while in my spare time. Since my wife and I just had a baby, I haven't had all that much spare time, but I thought I would outline my approach here to see if you all can keep me from making a(ny) stupid mistake(s). I have a nearly complete schematic, but I hesitate to post it, since going through a schematic is a lot to ask. I have a fair amount of experience, but not with audio, and not with mains-powered stuff.

The idea is to build a simple pre-amp with four stereo inputs, a volume control, and a balance control. Some other parameters are that there must not be any programmable parts (ROM's, uControllers, PLD's, etc.) and parts should not be too expensive or difficult to obtain. Ideally, everything would be stocked by Digikey. I would like the entire design to fit on a single circuit board with no wires running to any panel mount stuff.

If the design is any good at all, I'll post all the design info on the web somewhere so other people can build it if they want. I'm trying to avoid programmable parts to keep the entry barrier low. Plus, I don't have any kind of programmers at home, so it is convenient for me, too.

The power supply will use a standard AC receptacle (I'm in the US). I'll put a 5x20 mm fuse holder in series with the hot lead. Should I fuse neutral, too? There will be a mechanical pushbutton switch in series with the hot lead, also.

I'm using an 18 V 390 mA transformer (amvecco 70034, digikey TE70034-ND). I've got a bridge rectifier, and then an LM317 and an LM337 both in SOT223. I've selected resistors to generate +/- 15 V. There are also +/-

2.5V rails for logic. These are created from shunt regulators (TL431) from the +/- 15V.

I don't need caps across the bridge for such a small power supply, right?

Also, is this the right Voltage secondary for a +/- 15 V supply?

I've got 220 uF 63V electrolytic capacitors after the bridge. Does that sound right?

Would you put a fuse on the secondary side for this kind of thing?

There are four stereo inputs and a volume control. There is one output. I am aiming for a max gain of 1V/V.

Should I provide more or is that reasonable for consumer CD and DVD sources and typical amplifiers? This is not critical since I can easily add gain later by changing resistors.

The inputs are buffered with OPA2134's in the non-inverting configuration. There is a shunt resistor (10k ?) to set the input impedance. After the OPA2134's, I go to a 74HC4052 analog mux which is followed by another OPA2134 which then goes to a digital pot and finally to another OPA2134.

This seems like way too many op-amps. My rationale is that I want to keep the 4052 sandwiched between a low impedance source and a high impedance load. Also, the digital pot I am using, the AD5220BR10, can't stand input Voltages higher or lower than its rails (+/- 2.5) So I am putting a string of four diodes to and from ground on the digital pot input to keep the Voltage from exceeding the rails. The digital pot cannot withstand +/-15 V rails, so connecting it up there is not an option. Because of this low clipping level, I am keeping the pot sandwiched between op-amps, too, so I can attenuate the signal before the pot, and amplify it back up after, if necessary.

Channel select is done via a counter which is clocked by a user pushbutton (with debounce and Schmitt trigger). So the user can't directly select a single channel, but has to go through them sequentially. But hey, there are only four channels.

Volume up/down and balance left right is done with pushbuttons, too. The balance logic increments the digital potentiometer on one channel and decrements the other. The volume just increments or decrements both channels.

There are a lot of things about this design I don't like. The biggest two issues are:

1) I don't like having +/- 2.5V rails for the logic (I'd rather just drive it from 5V to ground.) 2) There are too many op-amps.

So, I am thinking of re-doing the volume. I could use the SSM2164. I would use the same potentiometers, followed by buffers to drive the control Voltage inputs on the SSM2164. I'd have to work out some way of doing balance, too, but I think that can be easily done so that one SSM2164 can do volume and balance for both channels. This would allow me to run from

+5V to gnd, instead of having split +/- 2.5V rails, get rid of the protection diodes, and lose a few op-amps in the signal path.

The other option for driving the SSM2164 would be to use a free running clock connected to a counter, connected to a DAC. The DAC output would drive the control input on the SSM2164. The volume up pushbutton would enable the counter to count up, and volume down would enable it to count down. Balance would count up on one channel and down on the other. If this is a good idea, I'd love to have a short list of inexpensive DAC's to look at. I could also build a "DAC" using 1% resistors directly from the counter outputs. (You know, 8 resistors in series from the lsb, 7 from the next bit, ... up to 1 from the msb). Resistors are very cheap, and I don't need great accuracy, although monotonicity would be nice to have for volume, and I doubt I could rely on monotonicity with such a set-up. I guess I could just do the math and see how bad such a setup could be.

And there would have to be some kind of logic to keep the counters from wrapping around, since the user would find that non-intuitive. (The digital pot already does this internally.) I should be able to handle the logic, but if anybody has some recommendation for a counter that makes it particularly easy, that would be great.

Thanks for reading my longish post.

--Mac

Reply to
Mac
Loading thread data ...

For some people, going through a schematic is a lot easier than going through so many words.

(snip)

No.

Are you talking about small ceramic caps across each diode? They are a good idea, because diodes can generate high frequency noise when they switch from on to off. You may not hear it in the audio, directly, but it might get into a nearby receiver. It is cheap insurance.

You could use a lower voltage transformer, say, the dual 15 volt one (and probably drop a size) that produces about 21 volts peak and 25 volt capacitors. This would lower the temperature of the regulators. The regulators just need a minimum of about 17 or 18 volts to accurately regulate 15, so 21 volts peak allows a ripple to dip 3 or 4 volts, including low line voltage and diode drops, before you drop out of regulation. The 12 volt version of the transformer will produce about 16, peak, but that isn't quite enough, unless you simplify to an unregulated +- 15 volt supply (which, depending on the schematic, is a might be a reasonable possibility). You should also probably put a .1uF film capacitor from each of the regulator outputs to ground to improve the high frequency regulation of the 317 and 337.

A bit light, perhaps, depending on the actual current drawn by the loads. You also need a pair of capacitors, one from each side of the bridge output to ground. With your transformer, I might use something more like a pair of 1000 uF @ 35 volts With a dual 15 volt transformer you could use 25 volt capacitors. 10% high line voltage would still drive them to only 23 volts peak.

I think I would put a pair of positive temperature coefficient resistors (that switch from a low resistance when cool to a very high resistance when hot) like MF-R020:

formatting link
between transformer and bridge. These will save your regulators from overheating if something gets shorted to ground, and they reset when they cool off.

Since various signal sources can vary a lot, I think you might need a gain of .1 to 10 for each input (or ganged for pairs of input channels).

I'm not going to think at that level of detail without seeing the schematic.

Why? And why 5 volts , total, since this limits the audio peaks to +-

2.5 volts, max. The CMOS chips can work with quite a bit more voltage than that. +- 5 or +-7.5 would seem more useful.
Reply to
John Popelish

...and then you go on to talk about DACs, digital pots, and whatnot.

How come?

What's wrong with the many time-proven simple designs based around analog components? Why are you reinventing the wheel?

I think you're making your life a lot harder than it needs to. If there are some sort of additional requirements that you didn't list, like a desire to learn new technologies or like a need to control it from a PC, it might be nice to know what they are.

I don't mean to be rude - I'm just suggesting that maybe there is a simpler answer than you're looking at.

Reply to
Walter Harley

In article , Mac wrote: [...]

No, never fuse the return line. In the US, the white and green wires are both at ground. The black is hot. You never want a failure mode that leaves the hot connected but opens the ground.

You want the fuse to be right at the entry point.

The regulators need capacitors on their inputs. These are effectively across the bridge so you end up with having put capacitors across the bridge.

You may also want to add some resistance between the bridge and the capacitors. This serves two purposes. It limits the peak current of the current from the bridge to the capacitors.

It also can reduce the risk of the regulators oscillating. If your wiring is a bit long and sloppy, it makes a whole bunch of Ls and Cs that can make tuned circuits with fairly high Q. If the regulator sees one of those it really likes, it will oscillate at that frequency. Adding the resistors lowers the Q. You could just use good wiring practice too to prevent the issue from coming up.

The peak to peak ripple is:

V = I/(120 C)

where: V is the peak to peak voltage (almost exactly) I is the load current in Amps C is the capacitance of a good capacitor in Farads.

You can figure out the ripple from that. If the lowest voltage into the regulator is enough to keep it working, you can use the "ripple rejection" value for the regulator to see what the output side ripple will be.

I'd go for a max gain of maybe 3.

It may not be enough op-amps. What are you hoping this thing will do?

So why didn't you do that and shift the DC point of the signals as needed?

Many of the up-down counters have a carry output that will work to indicate when another increment is a bad thing.

--
--
kensmith@rahul.net   forging knowledge
Reply to
Ken Smith

Reply to
DaveM

I hate wall warts. They are so big that they always take up more than one plug. And I have built small AC power-supplies before, but only for myself or for little projects back when I was in college. Since I am contemplating publicizing this design eventually I wanted to make sure that it was exposed to more expert scrutiny.

[snip]

I'm not sure if I understand this comment. The point of the buffers is to minimize the dependence of the switch resistance on the input Voltage (it is a CMOS switch). I must have mistakenly given you another impression.

I was planning on going DC-coupled.

[snip]

I will think about this.

I was afraid some people might not go to the trouble. But taking the time to visualize my description is troublesome, also, I guess. Anyway, thanks again.

--Mac

Reply to
Mac

Let's consider the PSU first.

Yes, a fuse in the ac live is essential. Neutral does not need fusing.

Secondary fusing is required by safety regs if the transformer isn't intrinsically safe.

You're going for 15V supplies, so the reservoir cap voltage mustn't drop below ~18V ( including ripple ) to ensure the regulation doesn't fail.

How do you plan to provide +/->18V to the regulators without using a centre tapped transformer ? I'll assume you mean 18-0-18 in that case.

Practical experience of various transformers shows that it is not easy to accurately predict the reservoir cap voltage. One of the larger issue tends to be how transformer regulation affects the final DC voltage. Some transformers have much higher copper losses than others. 18-0-18 sounds about right for starters though. Remember that the reservoir cap voltage also includes ripple voltage so you'll need more than 18V DC as measured on a meter.

The capacitance of the reservoir cap will influence both the pre-regulator ripple voltage and the ripple after the regulators.

Ripple voltage is determined by load current. What's your load current ?

Ripple voltage on 50Hz supplies is ~ 8.10^-3 x Iload / Cres. Substitute around

7.10^-3 for 60Hz . This from deltaV= t x I / C

Why use a 63V cap for the reservoir ? Use 35V and double the capacitance for the same can size / price.

Unless your load current is quite low I expect the SOT223 regulators will simply fry ! Why not use TO-220 ?

Graham

Reply to
Pooh Bear

The best will be to not have any mains AC on the board, it facilitates the handling immensly especially when you are not too experienced. Look at my light sculpture post on ABSE how to make a bipolar power supply with an AC adapter (in your case 15 or 18Vac single line).

You can just drop those buffers and directly switch the inputs with the mux, because the impedance change on the input is irrelevant since only one channel is active. Put 47k to gnd on each channel input, so the coupling caps are discharged and don't make pops. Put also a series resistor (1k is good) and then some diodes to the rails for protection. This digital pot is not good here, either you get a +/-15V version or make it with the ssm2164.

No, use a DAC or a simple linear pot for the ctrl voltage. Balance can be done by adding or subtracting a second voltage from another DAC or pot to the control voltage of right and the reverse to left, so only 2 sections of the 2164 are used. You could use the remaining sections to allow to mix two sources, fading from one source to another.

A circuit diagram would have been better than your life story.

--
ciao Ban
Bordighera, Italy
Reply to
Ban

dont forget to put 100nF caps accross the bridge rect diodes, Gets rid of a potential source of RF nasties

martin

Reply to
martin griffith

I reckon the issue is overstated. 1N400x diodes are so slow that it's not really a huge source of RF.

I do actually spec caps there now but I do it more as belt and braces for EMI conformity measurements as opposed to any intrusive audio effect.

Forgot to ask the OP if that's a toroidal transformer he's looking at btw.

Graham

Reply to
Pooh Bear

I know what the various colored wires are (in the US). Are you saying never fuse the white wire? I do know not to fuse the green wire (but I appreciate you emphasizing it).

Oh, yeah, I should have been more specific. I mean small anti-EMI caps across the bridge diodes.

Hmm. For some reason I was thinking that such a small transformer wouldn't need it. I don't think the short-circuit current is specified, but I could check it. Or I could just add some "opt" resistors in series with the secondaries.

I don't think there's going to be any stability problem.

In the US. People in 50 Hz environments need to use 100 C.

Well, I went through some such calculations in a spreadsheet. But depending on your assumptions about transformer regulation and best/worst case mains Voltage, it is not trivial (for me, at least) to get everything right.

Thanks.

Oh, you mean level shift using the op-amps. Duh. Thanks. I was thinking that I would have to shift by using a blocking cap (which I don't want to do), but that is not necessary. I can easily level-shift with the op-amps. I don't know why I didn't think of that. ;-)

Yeah. I just need to make sure I don't accidentally build a state machine where once max or min volume is attained the volume is no-longer adjustable. That wouldn't be too intuitive for the user, either. ;-)

--Mac

Reply to
Mac

How will a buffer make any difference to the switch resistance ? The applied voltage will be the same with or without a buffer ( ignoring minimal loading effects ).

Graham

p.s. a schematic would be appreciated

Reply to
Pooh Bear

I have often seen the 15V transformer turn out to be marginal, so I'd stay with the 18V.

The size of the (presently 220u) capacitor depends entirely on the load current. At 50mA load, 220uF is fine. I'd incrase it proportionately with load current - eg at 100mA use 470uF etc.

Roger Lascelles

Reply to
Roger Lascelles
[...]

Don't fuse the white wire. The odds of you getting a 3 prong outlet with the white and black inverted are very low. This is the only situation in which a fuse in the white would be an advantage. Normally, if the fuse on the white opens, the unit appears to be unpowered but in fact has deadly voltages inside.

They usually cause no harm so you can include them. An EMI filtering power entry module is also a good idea from that point of view. The down side of those entry modules is that they connect capacitors to the chassis ground. This is really a bad idea but has become standard practice.

You can measure the winding resistance with an ohmmeter and get an over estimate of the peak current. If you have an inductance bridge you can measure the leakage inductance. If not you could try a signal generator and scope. Short the secondary and measure the primary's impedance.

You need to consider the case when the switch is closed at the peak of the sine wave.

[...]

Confidence: The feeling you have before you understand the problem :)

Assume the worst number you came up with, if you are fairly sure there is no way to make it even worse. Remember an overdesign will work.

--
--
kensmith@rahul.net   forging knowledge
Reply to
Ken Smith

the

If it is fused properly they should not blow all the time... BTW, fuses blow when its most inconvenient, by definition ;o)

-- John

Reply to
John Smith
[snip]

I'm finding this to be pretty convincing. So I think I will take this advice. However, I will certainly check it out after it is built. I believe the 15 V transformer is pin-compatible, so I could always change it later. ;-)

--Mac

Reply to
Mac

(snip)

A much better question that you might answer first, is why are you using + and - 15 volts for the opamps and only +- 2.5 volts for the logic and CMOS switches? The while supply requirement is pretty arbitrary. You might do just fine running the opamps on +- 5, 7.5, 10 or 12 volts, depending on the rest of the design. As you say, you can pick a transformer later. You have more important questions to address first.

Reply to
John Popelish

[snip]

Point taken.

Yeah, OK. I'll put them in the schematic and layout. I may or may not install them. ;-)

Thank you. I am undecided on the 18 vs 15 Volt thing. I actually put 1uF electrolytic caps at the regulator inputs in the schematic but forgot to mention them (Another reason to post a schematic instead of describing the circuit). I could change them to 0.1 uF film capacitors. I will re-read the datasheets for the 317/337 and see what they say.

That's where these are.

Interesting. I think I will stick with the 18 Volts and see how it goes. I usually select capacitors with double the Voltage rating for the expected DC Voltage. That is why I'm using 63 V caps. Not sure where I picked that bit of baggage up.

I will double-check my calculations regarding capacitor size, too.

That seems a good idea. Thanks.

Wow. 10. OK. I'll use that as my starting point, and unless it seems preposterously loud with the gear I have, I'll keep it.

Fair enough. I'll post an ASCII art signal flow diagram somewhere in this thread. If you haven't lost interest, you can take a look. ;-)

[snip]
[snip]

The digital pot can only take 6 Volts (abs max) across its rails. I have pretty much decided not to use it directly for volume. I may use it instead to generate a control Voltage for a VCA (ssm2164).

I very much appreciate your detailed comments.

--Mac

Reply to
Mac

I now regret not just putting up a schematic.

Let me try to ASCII art the signal flow of the configuration as is (which I don't like), omitting numerous details:

input buffer A B |\\ | | |\\ | \\ -+-+-- | \\ | \\ | | | \\ in---+----|A=1+------------|mux |----|A=1+-------+ | | / | | | / | |\\ \\ | / ------ | / \\ | \\ R / |/ 74HC4052 |/ / | \\ TERM \\ \\

Reply to
Mac

Heh, heh. OK, I'll take another look at the datasheets and see if I can find any app notes.

That's true. Nothing exceeds like excess!

Thanks again.

--Mac

Reply to
Mac

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.